智能制造模式下基于改进BP-ARIMA组合模型产品需求预测方法

为有效预测智能制造模式下的不确定性需求,提出自回归移动平均模型ARIMA和改进BP神经网络的组合模型,对预测数据中包含线性规律的Lt以及非线性规律的ε_t进行模拟和分析,以解决预测有效性和精度问题.通过数据样本构建,对ARIMA模型结构进行辨识,确定p,d,q参数,并对模型进行诊断和检验;在此基础上进行需求数据一次预测;通过连接权值的修正降低BP神经网络学习误差,并对一次预测结果与原需求数据样本存在的误差进行二次预测.实例数据分析表明:组合模型的预测精度较ARIMA模型有显著提高,因此组合预测模型在预测效果上具有合理性和有效性.

Demand Prediction Method based on Improved BP-ARIMA Combination Model under Intelligent Manufacturing

Aiming to inhabit the bullwhip effect caused by uncertain demands under networked manufacturing (NM),we present a combined model composed of auto regressive integrated moving average mode (ARIMA) and improved back propagation (BP) neural network,which can simulate and analyze the linear rule Lt and non-linear rule εt included in predicted data.In this way,the effectiveness and accuracy could be improved.Through data sample construction,the structure of ARIMA model could be indentified and the further parameters p,d,q.After diagnostic test of ARIMA model,we accomplish the first prediction of demand.Furthermore,the learning error of BP neural network could be reduced according to weight modification.Thus,the second prediction for deviation prediction between original demand data and the first prediction result can be carried out.The real-world case simulation indicates that the accuracy of combined model is better than ARIMA model or BP,which also verify the effectiveness and rationality of presented model.